Beagle SCM

STATUS-009: status does not understand the track-vs-base split — a sub tracking its parent-repo pin is labelled trunk ?

The POST-026/027 track/base system (landed 59ba63c2, 2026-07-16) encodes both roles in the LAST get record of a wt's .be: the FRAGMENT is the base commit hash, everything else (scheme/host/path/query) is the TRACK — the followed branch/remote/worktree; the first row is the store anchor. So //TEST2/test#ec64900f… = tracking the parent wt's test gitlink, based at ec64900f. shared/wtlog.js:144 attachedBranch() — the ONE attach reader (DIS-057) feeding status's ?<branch> label, post's detach guard, and the ahead/behind divergence — predates the split: it only looks for a branch NAME in the query, so a track expressed as a remote/worktree URI reads as no-branch → trunk. Live in work/TEST2 2026-07-16: jab status prints ?\t93 ok / status test → ?\t360 ok — the sub tracks its pin as set in the parent repo, but status labels it trunk. Reported by gritzko (who also stated the record semantics above).

Input

Context

The attach record now carries track (non-fragment) + base (fragment); the reader only understands named-branch queries.

• work/TEST2/test/.be: 26716FVjFD get file:/home/gritzko/.be/?/beagle-ext-tests (store anchor), 26716FVjFE get //TEST2/test#ec64900f61417a263b7efd6b5dc847b4d288c595 (track = //TEST2/test, base = ec64900f…).
• Top-level .be: anchor file:/home/gritzko/.be/?/beagle-ext, last get ?#59ba63c2… (track = trunk, base = 59ba63c2…).
• jab status | cat → ?\t93 ok and status test → ?\t360 ok: empty branch token = trunk; the sub's actual track (the parent gitlink) is never shown.
• wtlog.js attachedBranch() derives attachment ONLY from a branch name in the record's query (refOf/parseBranch); a track that is a remote/worktree URI (//TEST2/test, an ssh: or file: source) has no query branch → branch="" → trunk. The base-in-fragment shape also grazes the DIS-075 detach test (query===undefined && isFullSha(fragment) gated on r.local) — a wrong-positive ATTACHED-to-trunk, not even detached.
• Blast radius beyond the label: views/status/status.js:296-300 (label), status.js:475 + JS-032 divergence (ahead/behind vs the WRONG ref), post's detach guard + curBranch — all route through attachedBranch() by design (DIS-057: "so they cannot disagree"); now they agree on the wrong answer.
• status.js:288 (SUBS-050) documents the label contract: named branch, or a mounted sub's /<title>/…/<branch>, or trunk ? — a URI-shaped track (parent pin, remote wt) has no rendering at all.

Goals

Spec ruled by gritzko 2026-07-16 (chat, transcribed): status's job is simpler now.

• The tracked target (branch, tag, parent pin, worktree, or remote) is read from the LAST get record in the wtlog; in get/post/patch records the #fragment is a commit hash and EVERYTHING ELSE is a ref to one of the above.
• Refs resolve to hashes ONLY through the unified resolver resolve_hash() (core/resolve_hash.js — RULE ZERO, meta resolve-hash law).
• Status computes ahead/behind the normal way: cur tip vs the resolve_hash()-resolved track tip.
• Status MUST show the tracked URI, and better also show the hash fragment; the shown fragment is the BASE hash — taken from the last get OR post record that has one — as an 8-hex hashlet.
• Repro-first: a wt whose last get record is <track>#<base-sha> with a URI-shaped track (TEST2 shape); assert the shown track URI + base fragment and the divergence ref; cover trunk-track and sub-tracks-parent-pin.
• The track/base record parse lives in ONE place — the reader (wtlog.js attachedBranch() or its successor), not each caller.
• DETACHED is exactly a bare-#sha last-get record (fragment, no ref) — gritzko 2026-07-16; the DIS-075 legacy special-cases in the reader reduce to this.

Constraints

The fix stays inside the one reader; no second attach parser.

• Ulogs contain only NORMALIZED URIs (gritzko 2026-07-16) — status shows the recorded track URI as-is; normalization is the WRITER's duty (get/post/patch), never a read-time re-canonicalization in status.

• DIS-057 ruling stands: status label, post detach-guard/curBranch, divergence ALL read attachedBranch() — extend it, do not fork per-caller logic.
• POST-026/027 are OPEN and still reshaping post's track/base writes — coordinate; do not hardcode today's row order if the spec says otherwise.
• Legacy single-row .be wts must keep reading exactly as before (DIS-075 detach forms included).

WIP

Design decisions

• (none yet)

TODOs

• repro: last get record <track>#<base-sha>, URI-shaped track → status prints trunk ?; assert shown track URI + base fragment + divergence ref
• the ONE reader returns {track ref, base sha}: track = last GET record minus fragment; base = fragment of the last get/post that has one
• resolve the track ref via resolve_hash() ONLY; ahead/behind vs that tip
• summary shows the tracked URI and the base hash fragment
• audit post's detach guard + curBranch and JS-032 divergence against the corrected reader
• legacy single-row .be regression: attach/detach forms of DIS-075 unchanged

Blockers and bummers

• POST-026/027 in flight — the track/base write shape may still move under this ticket.
• Side-find 2026-07-16: bare put vs the sub's unk repro file — RULED in PUT-009: mod-only bare put, unks stage manually; guard test rides this worktree.
• Side-find 2026-07-16: named delete now delegates into mounted subs (SUBS-055, the SUBS-039 put twin) — fix + tests ride this worktree too.

Outcome

• FILED 2026-07-16 off gritzko's live report in work/TEST2; nothing landed yet.